Build a nested resampling design without copying the data per outer fold

nested_resamples() builds the same nested resampling structure as rsample::nested_cv(), but stores index vectors into the original data instead of a materialized analysis set for every outer fold. For the same seed and the same specifications it produces the same splits; what changes is the size of the object that holds them.

Usage

nested_resamples(data, outside, inside)

Arguments

data: A data frame.
outside: The outer resampling specification, given either as an unevaluated call such as vfold_cv(v = 5) or as an already-evaluated rset object.
inside: The inner resampling specification, given as an unevaluated call such as vfold_cv(v = 5). Unlike outside, this cannot be an existing object, because it is evaluated once per outer fold.

Value

An object of class nested_resamples, which also carries the classes rsample::nested_cv() returns, so methods written against those keep working. It is the outer rset with an inner_resamples list column added, one inner rset per outer split.

Details

rsample::nested_cv() evaluates the inner specification against as.data.frame(split), so each outer fold's inner resamples reference their own copy of that fold's analysis set. Object size therefore grows by roughly one copy of the data for every outer fold. nested_resamples() evaluates the inner specification the same way, against the same transient frame, but keeps only the row indices it produces and remaps them onto the original data — so the inner splits reference the single shared copy the caller already has.

Differences from rsample

The splits select the same rows. rsample::analysis() and rsample::assessment() return identical frames, attributes included, and each inner split carries the class and the resample id rsample gives it, so labels() and rsample::add_resample_id() behave the same. What differs is what the splits point at: nestedtune's index the original data, rsample's index a materialized copy of each outer fold's analysis set. One behavior differs on purpose.

An outer bootstrap is refused, not warned about. The same observation can otherwise land in both the inner analysis and the inner assessment set, which makes the design invalid rather than merely unusual.

Examples

data(mtcars)

set.seed(1)
folds <- nested_resamples(
  mtcars,
  outside = rsample::vfold_cv(v = 3),
  inside = rsample::vfold_cv(v = 3)
)
folds
#> # Nested resampling:
#> #  outer: 3-fold cross-validation
#> #  inner: 3-fold cross-validation
#> # A tibble: 3 × 3
#>   splits          id    inner_resamples
#>   <list>          <chr> <list>         
#> 1 <split [21/11]> Fold1 <vfold [3 × 2]>
#> 2 <split [21/11]> Fold2 <vfold [3 × 2]>
#> 3 <split [22/10]> Fold3 <vfold [3 × 2]>

# Each element of inner_resamples is an ordinary rset.
folds$inner_resamples[[1]]
#> #  3-fold cross-validation 
#> # A tibble: 3 × 2
#>   splits         id   
#>   <list>         <chr>
#> 1 <split [14/7]> Fold1
#> 2 <split [14/7]> Fold2
#> 3 <split [14/7]> Fold3